[SR-3505] Optional of a wrapped type with value nil is incorrectly considered to be an Optional of another wrapped type

[swift-ci]

Previous ID	SR-3505
Radar	None
Original Reporter	goldsdad (JIRA User)
Type	Bug
Status	Resolved
Resolution	Done

Attachment: Download

Environment

OS X 10.11.6 (15G1212)
Xcode 8.2 (8C38)
IBM Swift Sandbox - Swift Ver. 3.0.2 (Release) - Platform: Linux (x86_64)

Additional Detail from JIRA

Votes	1
Component/s	Standard Library
Labels	Bug, TypeChecker
Assignee	@slavapestov
Priority	Medium

md5: e9915be07f69d31b7fe68a41773d1454

relates to:

• SR-6192 Spurious "Cast from ... to unrelated type ... always fails" warning when implicitly converting an empty collection

Issue Description:

The output from the following code is reasonable to me except for the last line producing "true".

enum Foo<T> {  
    case some(T)  
    case none  
}  
  
print(  Foo<Int>.some(3) is Foo<Int>               ) // output: true  
print(  Foo<Int>.some(3) is Foo<String>            ) // output: false  
  
print(  Foo<Int>.none is Foo<Int>                  ) // output: true  
print(  Foo<Int>.none is Foo<String>               ) // output: false  
  
print(  Optional<Int>.some(3) is Optional<Int>     ) // output: true  
print(  Optional<Int>.some(3) is Optional<String>  ) // output: false  
  
print(  Optional<Int>.none is Optional<Int>        ) // output: true  
print(  Optional<Int>.none is Optional<String>     ) // output: true  

The compiler even issues warning "Cast from 'Optional<Int>' to unrelated type 'Optional<String>' always fails" for the last line, as it does for the lines producing "false", yet the result is "true".

[jtbandes]

Dug into this a little, but I'm out of my depth. By looking at the generated IR, I saw that swift_dynamicCast is called for the Optional conversion/check, but a call is not made in the Foo case. It looks like as of @atrick's changes in #204 this might actually be expected behavior for Optional (there's a comment saying // Allow Optional<T>.none -> Optional<U>.none.) However I agree this is counterintuitive.

[belkadan]

@jckarter, we've seen this before, right?

[slavapestov]

I thought this was intentional, and the Sema warning is wrong.

[slavapestov]

#8876